How Does It Work

Sensor array processor designs are generally optimised, in terms of their implementational efficiency, for specific sensor array types, and invariably assume all the sensor array data to be "good" enough to form beams, which in operational environments is not always the case. The system overcomes this by decomposing the beamforming function into the following components :-
  • digital signal conditioning
    - band-shifting, filtering & downsampling of sensor array data - if appropriate;
    - automatic detection of malfunctioning sensors within the array and glitches within the sensor array data;
    - wideband compensation of malfunctioning sensors and bad data identified by the automatic detection routine;
    - optimal re-sampling of treated sensor array data;

  • management of data, address & coefficient stores
    - generation/updating of sensor position estimates - updating required for dynamic operation only;
    - generation/updating of beam steering address and coefficient stores - updating required for dynamic operation only;
    - updating of space-time store;

  • conventional beam formation
    - accession of address, coefficient & space-time stores;
    - generation of conventional time-domain beams at optimum sampling frequency, these referred to as broadband (BB) beams when run as a band-pass or multi-octave system, or wideband (WB) beams when run as a low-pass system.
In this way, the system is able to overcome the deficiencies and constraints of existing sensor array processor design, in an elegant fashion, incorporating as much intelligence as possible within the associated processing in order that a robust solution be obtained.

The property of fault tolerance, as addressed by the digital signal conditioning stage, is particularly relevant to the processing of digital data, where the amplitude of a signal from a malfunctioning sensor may default, with equal probability, to either the high state (with all bits set to 1) or the low state (with all bits set to zero). A single channel malfunctioning in the high state, for example, could result in the spatial information provided by the conventional beamformer being corrupted or possibly even destroyed.

Universality of JFilter SAP Beamformer Coefficients

The beamformer coefficients used for the conventional beam formation may be programmed to provide any combination of the following functions:-
  • beam steering - to steer an arbitrary number of beams, with arbitrary steering directions, yielding both azimuthal and elevational cover,
  • beam-dependent array shading - to enable use of arbitrary shading functions for each steering direction,
  • linear interpolation - to yield increased bearing accuracy,
  • range focusing - to maximise beamformer response to short-range sources,
  • base-banding - to enable efficient operation as either low-pass or band-pass system,
  • de-multiplexing - to correct for non-simultaneous sampling errors,
  • calibration - to correct for sensor amplitude and phase errors,
  • sensor data preservation - to enable production of conditioned sensor data or robust beam data &
  • periodic updating - to cater for time-varying changes to the positions of the sensors.
Given their applicability to sensor array systems of any size, shape and type, the beamformer coefficients may be regarded as being truly "universal", with the resulting beamforming function, in turn, being extremely versatile.

Mapping of JFilter SAP System Onto Different Platforms

The system comprises a suite of novel, hand-crafted processing algorithms, designed to yield high-throughput, reduced-complexity solutions, with the algorithms being coded in the "C" programming language for portability and optimum implementational efficiency. This enables the system to be ported onto any platform possessing a C/C++ compiler, and may thus be used as the basic building block for the construction of scalable, fault-tolerant digital sensor array processors, which may be efficiently implemented on either PC/Workstation or when assembled with commercial-off-the-shelf (COTS) DSP equipment.

Computational efficiency is achieved by means of a novel processing element (PE), referred to as the "Sino Computing Engine", the main component of which is a short fast Fourier transform (FFT) routine. The FFT, like the inner product operation required for carrying out the conventional beam formation, is a recursive algorithm which may be effectively exploited by most DSP devices. In addition, the system lends itself naturally to parallelisation, within each of its components, whilst the PE provides an upsampling/downsampling capability for optimisation of the sampling rates into, within and out of the beamformer.

The result is a digital sensor array processor, namely the JFilter SAP system, which may be regarded as a highly-parallel multi-rate DSP system, lending itself naturally to an efficient real-time implementation when assembled with COTS-based DSP equipment or when mapped onto application-specific integrated circuits (ASICs).

Accuracy, Flexibility and Ease of Use of JFilter SAP System

The fault tolerant nature of the system, achieved through the robustness of its design, combined with its versatility (being applicable to sensor arrays of arbitrary size and shape and for both broadband and narrowband systems), accuracy (in both spatial and temporal domains), flexibility (being able to steer arbitrary numbers of range-dependent, artefact-free beams in arbitrary directions) and ease of use, will benefit numerous application areas, including those of wireless (e.g. mobile and satellite) communications, medical imaging, sonar and radar (for both civil and defence), hearing aids & speech recognition, and geophysical & astrophysical exploration.